Multiscale Modelling of Materials Under Fusion Conditions

Read full text Summarise with VIRAT
Authors: Anirvinya Samanyu Tirumala, Felix Hofmann, Max Boleininger
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

We are currently facing a dilemma: our energy demands are continually increasing, while the urgency to reduce emissions due to climate concerns is more pressing than ever. Nuclear fusion presents a promising solution to this challenge, with the potential to provide clean electricity and minimal radioactive waste. However, one of the greatest obstacles in developing large-scale nuclear fusion reactors is designing materials capable of withstanding the extreme conditions within the reactor core. This study employs high-fidelity molecular dynamics (MD) simulations to model the microstructural evolution of materials under a wide range of irradiation conditions. While MD simulations generate high-dimensional datasets, the macroscopic properties relevant to engineering applications reside in a much lower-dimensional space. Although these simulations can be mapped to engineering-scale properties using characterisation algorithms, understanding how these properties evolve with simulation conditions remains a significant challenge. We present both a physics-informed and a data-driven approach to begin addressing this challenge. In the physics-informed framework, we parameterise a simplified model based on physical intuition, focusing on a single parameter: void content. The simplicity of this model facilitates straightforward upscaling to higher-level frameworks, such as finite element models (FEM), enabling practical engineering applications. However, extending this type of physics-based modelling to capture all macroscopic properties of interest is not feasible. Therefore, we adopt a data-driven approach, employing dimensionality reduction techniques and a vector autoregressive (VAR) model to investigate and characterise the evolution of the macroscopic properties.

Fields of Study & Subjects

Fusion materials and technologies Machine Learning in Materials Science Nuclear Materials and Properties

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
19 March 2026
Language
EN